Skip to content

Decay Segment Comparison

Comparative analysis — instantiates Activation Decay Measurement

Contrasts decay curves across cohorts to reveal who fades fastest and why, so timing is not set by a misleading population average.

Version
v1 · 2026-08-24 · History
Mechanism #
2441
Type
Comparative Analysis
Form family
Analysis, Modeling & Optimization
Solution family
Attention, Salience & Focus
Problem family
Observability, Measurement & Feedback Gaps
Problem subfamily
Temporal Sampling, Decay & Transition Resolution
Origin domain
Statistics & Experimental Design
Also from
Psychology
Instantiates
Activation Decay Measurement

A single population decay curve hides a dangerous assumption: that everyone forgets at the same rate. Decay Segment Comparison breaks that average apart. It builds a separate decay profile for each meaningful cohort — by role, channel, environment, or load — lays the curves side by side, and asks who fades fastest, and what explains the spread. Its defining move is contrast and diagnosis: it produces a comparison, not a schedule and not a cue. Crucially, it also watches the competing-cue noise each segment lives in — the rival stimuli that crowd out the primed state — because that noise is often the reason one cohort's curve plunges while another's holds. The output is an explanatory map of differential decay that later mechanisms can act on.

Example

A public-health authority runs a vaccination-reminder campaign and measures how long recall of "your booster is due" persists after the message. The blended curve says recall halves in about two weeks — but that average is a fiction. Segmenting, the analysis finds retirees who saw the message on a quiet local channel retain it for a month, while working parents reached through a crowded social feed lose it within four days. The comparison also logs the competing-cue noise: the parents' feed is saturated with rival notifications, and that noise tracks the steepness of their decay closely.

The deliverable is the contrast itself — a panel of cohort curves with the noise level annotated against each. It explains why the population average misleads (a fast-fading, high-noise segment dragging against a slow, quiet one) and hands that explanation on. It does not set anyone's reminder frequency or fire a cue; it diagnoses the spread.

How it works

  • Choose segmentation dimensions that plausibly drive decay — channel, role, load, environment.
  • Fit or estimate a decay profile per segment, not one blended curve.
  • Measure competing-cue noise per segment — the density of rival stimuli that erode the primed state — and correlate it with decay rate.
  • Rank and explain the spread, surfacing which cohorts fade fast and which factors travel with the fastest fade.

Tuning parameters

  • Segment resolution — coarse splits are robust but blur real differences; fine splits reveal them but thin the data and risk chasing noise.
  • Dimension selection — which attributes to segment on. The right dimension exposes the true driver; the wrong one hides it.
  • Noise-measurement fidelity — how carefully competing-cue density is quantified vs. estimated.
  • Significance guard — how large a cohort gap must be before it is treated as real rather than sampling scatter.
  • Explanatory depth — stopping at "who differs" vs. pressing on to "what factor explains it."

When it helps, and when it misleads

Its strength is catching the failure that a single curve conceals: a fast-decaying subgroup whose timing needs are erased when averaged into the whole, and the competing-cue noise that often causes it. It is the diagnostic that keeps population timing honest. Its failure mode is the aggregation trap in reverse — read only the blended curve and you can conclude the opposite of what holds for key subgroups, a Simpson's-paradox reversal where the aggregate hides or inverts the segment truth.[n1] The classic misuse is over-segmenting until random scatter looks like structure, "discovering" cohort differences that are pure noise. The guarding discipline is to segment only on dimensions with a mechanism behind them, require gaps to clear a significance guard, and always show the competing-cue noise alongside the curves so a difference has an explanation, not just a shape.

How it implements the components

  • cohort_specific_decay_profile — builds and compares a separate decay profile per segment, exposing the spread the average hides.
  • competing_cue_noise_monitor — measures the rival-stimulus density each segment lives in and ties it to that segment's decay rate as an explanation.

It diagnoses cohorts; it does not act on them: it does NOT implement refresh_or_reprime_protocol — turning a fast-fading cohort into a tighter refresh cadence is Refresh Cadence Adaptation, its cohort-side twin — and does NOT implement intervention_window_rule, which belongs to Activation Window Thresholding.

Editorial Notes

Form Classification

Form family: Analysis, Modeling & Optimization

Rationale: Decay Segment Comparison operates as a computation, comparison, model, or analytic representation used to infer, estimate, or choose because it contrasts decay curves across cohorts to reveal who fades fastest and why, so timing is not set by a misleading population average.

Independent corroboration: The frozen evidence defines Decay Segment Comparison as 'Contrasts decay curves across cohorts to reveal who fades fastest and why, so timing is not set by a misleading population average', so its operative form is Analysis, Modeling & Optimization.

Review outcome: Independent reviewer agreement; high confidence.

Origin Attribution

Primary origin: Statistics & Experimental Design

Origin pattern: Cross-disciplinary synthesis

Present-day reach: Multi-domain

Rationale: Stratified statistical analysis cohered fitting and comparing group-specific trajectories to reveal heterogeneous decay hidden by a population average and to diagnose interaction with segment conditions.

Related originating lineages:

  • Psychology — Memory and priming research supplied competing-cue interference as a substantive explanation for differential fading.

Review resolution: Stratified statistical analysis cohered fitting and comparing group-specific trajectories to reveal heterogeneous decay hidden by a population average and to diagnose interaction with segment conditions.

Review outcome: Reconciled after independent review; high confidence.

Notes

[n1] Simpson's paradox is the phenomenon where a trend that appears in aggregated data disappears or reverses once the data are split into meaningful subgroups. It is the formal warning behind reading only a blended decay curve.